Re: [RFA] _Jv_CompileMethod

[Bryce McKinlay <mckinlay at redhat dot com>]

Hi Keith,

Why call _Jv_InterpMethod::run() instead of _Jv_InterpMethod::compile() 
directly? That would make more logical sense I think, and the 
compile_mutex locking could be moved into compile() in order to handle 
the case where a request to compile were made simultaneously by both the 
interpreter and the JDWP.

Also, for the current patch, it looks like the "When args is NULL, don't 
run anything -- just compile it." comment is on the wrong method.

Do you have commit access for gcc/libgcj?

Bryce


Keith Seitz wrote:
> Hi,
>
> The following patch adds support to _Jv_InterpMethod to allow an 
> outside caller (like JDWP support) to compile methods.
>
> This is necessary because the line tables are adjusted when the method 
> is compiled. If the debugger happens to request a line table for a 
> method before it is compiled, it will have a different line table than 
> the VM.
>
> Keith
>
> ChangeLog
> 2006-01-18  Keith Seitz  <keiths@redhat.com>
>
>         * include/java-interp.h (_Jv_CompileMethod): Add declaration.
>         (class _Jv_InterpMethod): Add _Jv_CompileMethod as a friend.
>         * interpret.cc (_Jv_CompileMethod): New function.
>         (run):  Massage code to allow for NULL args.
>         Update comments to explain NULL args.
>         Return if compiling the method and args is NULL.